强化学习范式是一种基于奖励的学习算法，与 BionicSoftHand 配合使用。这意味着，无需定义机器人必须模仿的特定动作，而只需给手一个目标。然后尝试通过反复试验来实现这一目标。根据收到的反馈（包括肯定和否定），它会逐步优化其操作，直到最终成功完成任务。
具体来说，BionicSoftHand 的任务是旋转十二面的立方体，以使先前定义的侧面指向端部。使用数字孪生体体在虚拟环境中学习必要的运动策略，该数字孪生体体是用深度摄像头和人工智能算法提供的数据创建。
数字仿真模型可以极大地加速学习，尤其在您复制这个模型时。通过大规模并行学习，所获得的知识将与所有虚拟手共享，然后虚拟手将继续使用新的知识 — 这意味着每个错误只会发生一次。成功的工作可立即用于所有模型。
在仿真中对控制器进行训练后，会将其传输到实际的 BionicSoftHand。使用在虚拟环境中获取的移动策略，可以将多面体转到指定侧，并在将来相应地移动其他物体。因此，可以将已学习的知识和新技能的基本构成要素与其他机械手共享，并在全局范围内使用。
Unlike the human hand, the BionicSoftHand does not have any bones. It controls its movements via the pneumatic bellows structures in its fingers. When the chambers are filled with air, the fingers bend. If the air chambers are empty, the fingers remain stretched. The thumb and index finger are additionally equipped with a swivel module, which allows these two fingers to also be moved laterally. This gives the bionic robot hand a total of twelve degrees of freedom.
The bellows in the fingers are enclosed in a special 3D textile cover knitted from both elastic and high-strength fibres. This means that the textile can be used to exactly determine at which points the structure expands, thereby generating force, and where it is prevented from expanding.
In order to keep the amount of tubing work required for the BionicSoftHand as low as possible, the developers have specially designed a small, digitally controlled valve terminal, which is mounted directly below the hand. This means that the tubes for controlling the fingers do not have to be pulled through the entire robot arm. As a result, the BionicSoftHand can be quickly and easily connected and operated with only one tube each for supply air and exhaust air. The proportional piezo valves used enable the movements of the fingers to be precisely controlled.